DISCUSSION PAPER. The Welfare Effects of Fuel Conservation Policies in the Indian Car Market. R a n d y C h u g h a n d M a u r e e n Cr o p p e r

Transcription

1 DISCUSSION PAPER September 2014 RFF DP The Welfare Effects of Fuel Conservation Policies in the Indian Car Market R a n d y C h u g h a n d M a u r e e n Cr o p p e r 1616 P St. NW Washington, DC

2 The Welfare Effects of Fuel Conservation Policies in the Indian Car Market Randy Chugh and Maureen Cropper Abstract We estimate a model of vehicle choice and miles driven to analyze the impact of fuel conservation policies in the Indian car market. Taxing diesel fuel to equalize diesel and petrol prices would reduce fuel consumption in the new car market by 7 percent and reduce diesel car sales by 26 percent. A tax on diesel cars with the same sales impact would reduce fuel consumption by only 2 percent. The compensating variation per liter of fuel saved is smaller for the fuel tax than for the car tax; however, the car tax has lower deadweight loss per liter of fuel saved. Our estimates of the long-run elasticities of fuel consumption with respect to fuel prices imply that the CAFE standards contemplated by the Indian government would generate a significant rebound effect. Projected fuel savings are 20 percent if consumers do not adjust to the change in operating costs and less than 9 percent once consumers adjust. Key Words: Indian car market, fuel conservation, fuel taxes JEL Classification Numbers: L9, R48, Q Resources for the Future. All rights reserved. No portion of this paper may be reproduced without permission of the authors. Discussion papers are research materials circulated by their authors for purposes of information and discussion. They have not necessarily undergone formal peer review.

3 Contents 1. Introduction Overview of the Indian Car Market, Fuel Pricing, and Fuel Consumption A Discrete-Continuous Choice Model of New Car Purchases The Model Vehicle Choice Driving Distance Estimation Strategy Data and Empirical Specification New Car Buyers and Vehicles Purchased Model Specification Econometric Results Policy Simulations Results Discussion Implications of Our Results for CAFE Standards Conclusions Tables and Figures References Appendix: Optimal Externality Taxes... 34

4 The Welfare Effects of Fuel Conservation Policies in the Indian Car Market Randy Chugh and Maureen Cropper 1. Introduction The Indian car market has grown rapidly since the economic reforms of the early 1990s. Recent years have seen a doubling of new car sales between 2002 and 2006 and again between 2006 and 2011 (SIAM, various years). The passenger vehicle fleet, which stood at 22 million vehicles in 2010, is conservatively projected to increase to 112 million vehicles by 2030 (Arora et al. 2011). Growth in the car market has brought with it increased foreign oil dependence, fuel consumption, and associated externalities. The Indian government has responded to local air pollution concerns with fuel quality and emissions standards. However, at the center of the policy debate in a country that imports 80 percent of its oil is what policies should be used to reduce fuel consumption. The Indian government is currently contemplating fuel economy standards and, since April 2013, has required fuel efficiency (kilometers per liter [km/l]) information to be posted on all new cars. At the same time, India taxes diesel fuel at a lower rate than gasoline (petrol). Between 2002 and 2013, petrol sold for 30 percent to 50 percent more than diesel. The gap between diesel and petrol prices, shown in Figure 1, has amplified the fuel economy advantage of diesel cars and led to a sharp rise in their sales. The diesel share of the new car market rose from 22 percent in 2002 to 34 percent in As we document below, diesel buyers, on average, drive farther than buyers of petrol vehicles. Thus dieselization has increased the fuel consumption associated with a rapidly expanding fleet. To curb fuel consumption, the Indian government is now contemplating fuel economy standards, despite the potential efficiency of taxing diesel fuel instead. The government is also considering a tax on diesel vehicles, which is more politically feasible than taxing diesel fuel. Chugh, Department of Justice; Cropper, University of Maryland and Resources for the Future; cropper@rff.org. We thank Resources for the Future and the World Bank for funding. We also thank seminar participants at the University of Virginia, Cornell University, the National Bureau of Economic Research (NBER), and Stanford University for helpful comments and Marisol Rodriguez Chatruc for excellent research assistance. The views expressed herein are entirely those of the authors and should not be presumed to reflect those of the US Department of Justice. 1

5 The 2010 report from the Expert Group on a Viable and Sustainable System of Pricing of Petroleum Products (Parikh 2010) considered equalizing the prices of diesel and petrol fuel but instead recommended that an additional excise duty on a diesel vehicle corresponding to the differential tax on the petrol should be levied. This paper compares the welfare implications of the diesel car tax recommended by the Expert Group to a diesel fuel tax. We present a structural econometric analysis of the market for new cars and simulations of market responses to these alternative policies. Using data from the J. D. Power APEAL survey, we model the joint decision of which car to buy and how much to drive it in a mixed logit discrete-continuous choice framework. We estimate the model for the year 2010 and simulate consumers responses to three fuel conservation policies: a diesel fuel tax that would equalize the prices of petrol and diesel fuel, a diesel car tax that would result in the same reduction in the market share of diesel cars, and a smaller diesel fuel tax that would result in the same total fuel savings as the diesel car tax. For each policy, we simulate changes in market shares, driving distances, and total fuel use. We compare the efficiency of policies by calculating compensating variation, government revenue, and deadweight loss per liter of fuel conserved. Our results quantify the efficiency of the fuel tax relative to the car tax. In 2010, a 34 percent diesel fuel tax would eliminate the disparity between diesel and petrol prices and reduce the market share of diesel cars from 34 percent to 25 percent. A diesel car tax of 25 percent would achieve the same reduction in market share. The fuel tax, which lowers the miles driven by diesel car buyers and shifts would-be diesel buyers to petrol cars, reduces total fuel consumption in the new car market by 7 percent. The car tax, which operates only by shifting diesel buyers into petrol cars, reduces total fuel consumption by only 2 percent. 1 Compensating variation, the amount new car buyers would have to be given to restore them to their prepolicy level of utility, is almost twice as large per liter of fuel conserved for the car tax as for the fuel tax (120 versus 64 rupees per liter [Rs/L]). 2 1 Raising the tax on diesel fuel or taxing diesel cars would have effects outside of the market for new cars. A tax on new diesel cars would increase the lifetimes of used diesel cars. The diesel fuel tax would alter the price of used diesel cars relative to used petrol cars and would reduce fuel consumption by owners of existing diesel cars. 2 A more policy-relevant comparison from the perspective of fuel conservation is to compare the diesel car tax with the diesel fuel tax that achieves the same total fuel conservation. A diesel fuel tax of 7.6% would achieve the same reduction in petrol plus diesel fuel consumed by new cars. The compensating variation of the smaller fuel tax is 58 Rs/L. 2

6 The deadweight loss of the car tax is, however, lower per liter of fuel conserved in fact, it is negative ( 27 Rs/L for the car tax versus 12 Rs/L for the diesel fuel tax). The negative deadweight loss per liter of fuel conserved occurs because buyers who switch from a diesel to a petrol car generate enough government revenue through their consumption of petrol fuel to more than offset the welfare impact of the diesel car tax. Effectively, buyers are shifting from a market with a lower tax (diesel) to a market with a higher tax (petrol). The negative deadweight loss associated with the diesel car tax is possible in the case of multiple distorted markets (Goulder and Williams 2003) and should be interpreted as decreasing preexisting deadweight loss in the market for petrol. Whether the deadweight loss is policy (and politically) relevant depends on whether the revenues from the taxes we consider are recycled. In principle, revenues from either the diesel fuel or car taxes, and the increased petrol revenues, could be used to reduce the Indian income tax rates. If revenues are not recycled, the compensating variation of the large diesel tax is about 3.0 percent of average diesel car buyers incomes; the compensating variation of the car tax is about 1.6 percent. Our analysis informs fuel conservation policy in several ways. In addition to pointing out the efficiency costs of each policy and its revenue-raising implications, we trace out the implications of diesel fuel and car taxes for future fuel consumption. Given the rapid growth in the Indian car market, enacting either of these policies today would have a significant cumulative effect on fuel conservation as well as on the composition of the passenger vehicle fleet. Using conservative projections of the growth in the Indian car market (Arora et al. 2011), the magnitude of fuel conservation achieved in 2030 from enacting either policy today would be 40 times as great as the reduction today. This emphasizes the benefits of enacting sensible policies when the car market is in its infancy. Our results also have implications for the CAFE standards currently being contemplated by the Indian government. The Bureau of Economic Efficiency is contemplating weight-based corporate average fuel economy (CAFE) standards that would go into effect in , with more stringent standards proposed for (Roychowdhury 2011). Our estimates of the long-run elasticity of fuel consumption with respect to fuel price suggest that improving vehicle fuel economy is likely to have a significant rebound effect. As we demonstrate, the proposed fuel economy standards would reduce fuel consumption by approximately 20 percent if consumers continued to purchase the same vehicles they bought in 2010 and drove them the same number of miles. Our model suggests that allowing for both types of adjustments implies a reduction in fuel consumption of only 8.5 percent. 3

7 The rest of the paper is organized as follows. Section 2 describes the new car market, fuel pricing, and fuel consumption in India. Section 3 presents our model of vehicle choice and miles driven and our estimation strategy. Section 4 looks at the data used to estimate the model, including the stylized facts about Indian cars and the people who buy them. Section 5 gives our estimation results. Section 6 discusses the results of policy simulations, and section 7 concludes. 2. Overview of the Indian Car Market, Fuel Pricing, and Fuel Consumption Sales of passenger vehicles in India have been growing rapidly, from 600,000 cars in 2002 to 1.2 million in 2006 and 2.6 million in 2011 (SIAM, various years). Hatchbacks constitute approximately two-thirds of new car sales, with sedans accounting for about 17 percent of the market and sport utility vehicles (SUVs) and multiuse vehicles (MUVs) the remainder. Trends in sales of diesel passenger vehicles in the Indian car market are shown in Figure 2. Diesel vehicles accounted for 34 percent of new car sales in 2010, although there was significant variation across vehicle types. As relatively few diesel hatchbacks are available because of technological constraints, 3 diesel models share of the hatchback market has remained between 10 percent and 20 percent. Diesel models share of India s biggest cars, SUVs and MUVs, has also been relatively constant, between 60 percent and 70 percent. In the sedan market, however, the share of diesel cars has increased from 25 percent in 2003 to nearly 50 percent in The increasing trend in purchases of diesel vehicles can be explained in part by the lower price of diesel fuel. Figure 1 shows the retail prices of diesel and petrol in Delhi from 2001 to the present. Prior to 2010, the year of our study, both diesel and petrol prices were governmentdetermined and, as the figure suggests, shielded from variation in world oil prices. The base price received by oil companies was set by the federal government. 4 To this were added customs and excise duties and sales taxes. Beginning in 2010, the base price of petrol was allowed to vary with international oil prices. On average, the price per liter of diesel was about 70 percent of the price of petrol, although the gap between petrol and diesel prices increased after the base price of petrol was allowed to vary with international oil prices. 3 There are no diesel engines below 1,250 cubic centimeters (cc), but many hatchbacks have smaller engines. 4 The base price was set equal to 80% of the import price of oil plus 20% of the export price (Anand 2012). 4

8 The lower price of diesel in Figure 1 reflects lower taxes on diesel and, in some years, a discount in the price retail dealers were charged for diesel. For modeling purposes, we treat the difference between petrol and diesel prices as a difference in tax rates. 5 An important question is why diesel is taxed at a lower rate than petrol. Sixty percent of diesel fuel is used for road transport, primarily for trucks and buses. Approximately 20 percent is used for power generation (both captive power generation and transmission to the grid), 12 percent by agriculture, 4 percent by railways, and 4 percent for miscellaneous uses (Anand 2012). Diesel is widely perceived to be a poor man s fuel, and there is concern about the macroeconomic consequences of equalizing the prices of diesel and petrol (Anand 2012; Parikh 2010). For these reasons, there are political pressures not to raise the tax on diesel fuel, but to tax diesel cars instead. Why are people buying diesel cars? As discussed below, we find operating cost to be a key determinant of vehicle choice. Although diesel cars generally are more expensive than their petrol twins, their lower operating cost more than offsets the purchase price difference (Chugh et al. 2011). The fuel economy of a diesel sedan (about 14.4 km/l in 2010) is about 22 percent higher than that of a petrol sedan (11.8 km/l) (see Table 1). When coupled with the 30 percent cheaper price of diesel fuel, the diesel sedan s fuel economy advantage results in an operating cost that is approximately 60 percent that of a petrol sedan. In view of the lower operating cost of diesel vehicles, it is not surprising that they are driven more. In 2010, diesel sedans were driven 36 percent farther than petrol sedans, diesel SUVs were driven 58 percent farther than petrol SUVs, and diesel hatchbacks were driven 66 percent farther than petrol hatchbacks (see Table 2). 3. A Discrete-Continuous Choice Model of New Car Purchases We model the purchase and use of new cars in a discrete-continuous choice framework. The method, pioneered by Dubin and McFadden (1984), provides a tractable, theoretically motivated approach to dealing with selection bias and has become a workhorse model in energy demand estimation. The key insight of their study is that if consumers with high expected electricity usage buy low-operating-cost appliances, a simple regression of usage on operating cost will result in a biased estimate of the price responsiveness of electricity demand. By directly 5 The discount to retailers of diesel, referred to as underrecovery, is essentially a subsidy from the federal government to diesel purchasers. Oil companies sell diesel fuel to dealers at a discount and are then compensated by the federal government. It is currently the case that the taxes on diesel, minus the underrecovery, are positive. 5

9 modeling the discrete choice of which appliance to purchase, the authors develop a selection correction method and recover unbiased elasticity estimates in a second stage. This two-stage approach has been applied to the US car market in several studies. Goldberg (1998) uses a model of vehicle choice and utilization, coupled with an oligpolistic model of supply, to study the effect of CAFE standards on car sales, prices, and fuel consumption. West (2004) follows a similar approach, considering a broader range of policies and studying their distributional effects. One drawback of the two-stage approach is that separate estimation of car choice and miles driven leads to two sets of model parameters, often differing in magnitude and sign. As the number of miles driven is derived using Roy s identity in a static utility maximization framework, theoretical consistency requires a single set of parameters to determine both choices. This is especially important in calculating the welfare impact of policy interventions. Recent contributions from Feng et al. (2013) and Bento et al. (2009) have sought to overcome this limitation by introducing simultaneous estimation techniques, which we follow here. 6 Our approach incorporates these recent modeling and estimation advances in a mixed logit, discrete-continuous choice model of which car to buy and how much to drive it. We incorporate body type and manufacturer fixed effects to account for unobserved vehicle characteristics, and we allow for randomly distributed parameters to account for unobserved household characteristics. The model is estimated by full information maximum likelihood, which leads to a single set of parameter estimates, allowing for theoretically consistent welfare estimates The Model The household s decision takes the form of a standard static utility maximization problem where utility is a function of car characteristics, kilometers driven, and consumption of all other goods. The household chooses the car that yields the highest indirect utility; optimal driving distance can then be inferred by Roy s identity. 6 We note two recent studies of vehicle demand in China (Li 2014; Xiao and Ju 2014) that use aggregate household data to analyze the impact of policies to limit vehicle emissions. Fullerton et al. (2014) estimate a model of vehicle demand and miles driven for Japan using aggregate data. We are fortunate to have individual household data on miles traveled and vehicle choice. 6

10 Although the J. D. Power survey is conducted in several locations across the country, we model the new car market as a single national market with the choice set being the same for all households. As data are limited to households that have purchased a new car in the survey year, the choice set does not include an outside good. Thus households in the model are faced with the decision of which car to buy conditional on already having decided to buy a new car. This modeling approach is necessary given data limitations, but it also allows for a more precise estimation of means and distributions of preferences for the subpopulation of new car buyers (for further discussion, see Train and Winston 2007) Vehicle Choice Each household chooses the car from choice set that yields the highest utility. Following Bento et al. (2009), household s utility conditional on buying car is ( ), (1) where is annual income of household minus the annualized rental cost of car ; is a vector of characteristics of car, characteristics of household, and interactions of the two; is the per-kilometer operating cost of car ; is an idiosyncratic taste for driving; and is an independent and identically distributed (i.i.d.) stochastic preference shock. 7 The coefficients and are assumed to follow uncorrelated random distributions, the parameters of which are estimated along with other parameters of the model. For example,, where is the sample mean of and is an idiosyncratic deviation drawn from some distribution. 8 The idiosyncratic taste for driving,, is assumed to be normally distributed with mean zero; its standard deviation,, is estimated along with the other parameters of the model. Let represent the common set of coefficients such that. Individual parameters are then distributed according to the joint probability density function 7 This functional form leads to a log-linear specification of the demand for kilometers driven. Previous discretecontinuous models, including Dubin and McFadden (1984), Goldberg (1998) and West (2004), use an indirect utility function that leads to a linear demand function for kilometers driven. We follow the more recent literature in making use of the functional form in equation (1). 8 Including random coefficients on all car characteristics would result in a more general model, but at the cost of a higher dimensional integral requiring many more random draws to simulate. Experiments with more general specifications did not improve model fit or substantially change counterfactual predictions. 7

11 . Assuming that have a Type I Extreme Value distribution, the probability that household chooses car takes the mixed logit form, (2) where is the scale parameter of the i.i.d. type I extreme value error term Driving Distance Using Roy s identity, annual driving distance can be derived from equation (1) as follows: ( ). (3) As mentioned above, household s idiosyncratic taste for driving,, is drawn from a mean-zero normal distribution with standard deviation, which is estimated along with the other parameters in the model. This modeling of the driving distance decision, which follows Feng et al. (2013), improves the fit of the model to the data but still allows for a more general correlation of errors as in Bento et al. (2009). Taking account of the fact that the same randomly distributed coefficients that determine vehicle choice probabilities also determine driving distance predictions, the log of demand for kilometers driven equation becomes ( ) (4) Just as equation (2) takes advantage of the closed form solution of the integral over the type I extreme value preference shock, equation (5) below takes advantage of the closed form probability density function for the normally distributed idiosyncratic taste for driving. Thus the likelihood of observing kilometers driven conditional on household buying car is ( ) ( ) (5) where is an indicator function equal to 1 if household bought car and 0 otherwise, and is equation (4) without the idiosyncratic taste shock. 8

12 3.4. Estimation Strategy Household s likelihood of buying the car it is observed to have bought and driving the distance it is observed to have driven is the product of the probability of buying car (equation (2)) and its likelihood of driving conditional on buying car (equation (5)). The full information likelihood function is the product over all households, (6) The log-likelihood function is (7) Evaluating the log-likelihood function directly would require solving the integrals in equations (2) and (4). In the absence of closed-form solutions, integration can be performed by simulation (Train 2009). For any draw from the distribution, the log-likelihood for household is calculated, the sum of the log-likelihoods from separate draws is found, and the average is taken. In the limit as approaches infinity, simulation error approaches zero. The second departure from equation (7) is to weight each observation to ensure that the prevalence of each model in the sample is proportional to its prevalence in the market. 9 Thus the log-likelihood to be maximized is given by (8) where is the weight applied to observation and probabilities have been replaced by simulated values Weights for each observation equal the ratio of the market share to the sample share of the chosen model. 10 Results presented below are based on integrals simulated using 200 shifted and shuffled Halton draws, a quasirandom scheme that provides better coverage than pseudorandom draws. While some studies use up to 5,000 pseudo-random draws, Train and Winston (2007) find 200 Halton draws to be sufficient. We follow their approach to testing for sufficient draws by calculating the value of the test statistic using 400 draws at the parameter estimates obtained using 200 draws. Under the null hypothesis that the gradient is zero, this test statistic is distributed chi-squared with degrees of freedom equal to the number of parameters. Using this approach, we fail to reject the hypothesis that the parameters found using 200 draws are indeed likelihood maximizing. As in Train and Winston (2007), we present standard errors that are robust to simulation noise. 9

13 4. Data and Empirical Specification 4.1. New Car Buyers and Vehicles Purchased We estimate the model using data on household car choice and monthly driving distances from the 2010 J. D. Power APEAL survey of 7,000 new car buyers in India. The survey provides the make, model, and fuel type of the car purchased; the purchase price; monthly kilometers driven; and the buyer s estimate of fuel economy. It also collects data on household income and demographic characteristics and information on vehicle ownership. Car characteristics data come from the magazine AutoCar India and the website DriveInside.com. Most car models are available in multiple versions (e.g., a Honda Civic LX or EX). This level of detail is available in AutoCar India and on DriveInside, but survey respondents report a model/fuel type only. Car characteristics for each model/fuel type are constructed as the unweighted average across all model versions of each fuel type. Table 1 presents the sales-weighted summary statistics for all vehicle models sold in Price and fuel economy variables are taken as the average across all respondents for each model/fuel type but are found to be similar to price and fuel economy reported in AutoCar India. 11 It is important to note that within-body type differences across fuel types are influenced by the difference in model availability across fuel types. Of the 54 models in the dataset, 31 are available in both petrol and diesel form. Among these models, 20 were bought in substantial numbers in both petrol and diesel form, while 11 were purchased as diesels only. Counting only those vehicles that sold in substantial numbers yields a choice set of 74 model/fuel types for each buyer. For hatchbacks and sedans, every diesel model is available as a petrol vehicle, but a wide variety of petrol models are available for which there is no diesel counterpart. Thus the differences between petrol and diesel vehicles observed in Table 1 reflect model availability in addition to differences between diesel and petrol twins. Nevertheless, some stylized facts are worth noting. Diesel cars are heavier than petrol cars, have a lower horsepower-to-weight ratio, and, with the exception of SUVs, have higher fuel economy. 11 A regression through the origin of buyers estimates of fuel economy on published estimates of city fuel economy yields a coefficient of 1.14 (s.e. = 0.010). When highway fuel economy is added to the equation, the coefficient on city fuel economy equals 1.09 (s.e. = 0.099) and the coefficient on highway fuel economy is (s.e. = 0.071). 10

14 Because of their higher fuel economy and cheaper fuel, diesel cars have lower operating costs across body types. 12 In 2010, diesel operating costs were 34 percent lower for hatchbacks, 39 percent lower for sedans, and 20 percent lower for SUVs. Predictably, the owners of loweroperating-cost vehicles drove more. The APEAL survey provides information on the income, age, gender, family size, and car ownership of respondents. Sedan owners, on average, have higher incomes than hatchback owners. Diesel sedan owners have, on average, lower incomes than petrol sedan owners. Family size is slightly higher among diesel households, and the average age of diesel car owners is slightly lower. Family size is correlated with vehicle size: for both diesel and petrol vehicles, family size is smaller on average for hatchback buyers than for sedan buyers and smaller for sedan buyers than for buyers of SUVs Model Specification To operationalize the model, it is necessary to convert the purchase price of a vehicle to an annualized rental price and to construct a per-kilometer operating cost. We focus entirely on the purchase price (inclusive of sales taxes) and calculate the rental price as the annual payment on a car loan such that the loan would be paid back over the expected life of the vehicle. Vehicle survival probabilities are based on a survival curve for Indian cars estimated by Arora et al. (2011). Their survival curve assumes a maximum vehicle life of 20 years and implies an expected vehicle life of 18 years. We use a nominal interest rate of 15 percent, based on interest rates charged on new car loans in India and note that about 80 percent of new car purchases are financed with such loans (Carazoo 2011; Seth 2009; Shankar 2007). After adjusting for inflation, we use a real interest rate of 8.5 percent. Operating cost is the fuel price in Delhi divided by fuel economy. 13 As with vehicle price, fuel economy is taken as the average self-reported fuel economy for each vehicle type, but results are robust to the use of AutoCar India data vs. survey data. The Delhi prices of petrol and diesel fuel in 2010 were 49.4 and 36.7 Rs per liter, respectively. Our vector of vehicle characteristics includes all of the attributes listed in Table 1, make dummies, and dummies for 12 Operating costs are calculated using Delhi fuel prices, as described more thoroughly below. 13 The variation across cities in diesel and petrol fuel prices is small. In 2010, the average price of petrol in 31 Indian cities was Rs with a standard deviation of 2.40 Rs. The average price of diesel was Rs with a standard deviation of 1.74 Rs. 11

15 body type (sedan and hatchback; SUV is the omitted category). Vehicle attributes include performance characteristics (engine size, torque, and the ratio of horsepower to weight), number of gears, whether the vehicle has an automatic transmission, and measures of vehicle size (length, width, height, ground clearance, and weight). We also include indexes of the safety features and the luxury features of the vehicle. 14 To improve model fit and better characterize substitution possibilities, we interact vehicle and household characteristics and allow three of the coefficients in the indirect utility function to be random. We interact family size with sedan and hatchback dummies and with the ratio of horsepower to weight (power ratio). We interact buyer age with safety index. The distribution of the income minus rental cost coefficient is assumed to be log-normal to reflect the positive marginal utility of consumption of all other goods and the positive wealth effect on driving distance, with. Following the same reasoning, the distribution of the operating cost coefficient is assumed to be negative log-normal such that. with 5. Econometric Results Figures 3 and 4 summarize the within-sample fit of the estimated models in terms of market shares and annual kilometers driven. Aggregated to body- and fuel-type categories, predicted market shares match actual market shares closely. In fact, the model mirrors the within-sample shares of petrol and diesel vehicles (67.4 percent and 32.6 percent, respectively) to two significant digits. On average, the model overpredicts fuel usage and kilometers driven by 4.75 percent. Annual kilometers driven are predicted accurately for four of the six body-fuel type categories but are overpredicted for drivers of petrol hatchbacks and underpredicted for owners of diesel SUVs. Table 3 presents estimation results for all parameters; manufacturer fixed effects are not shown. Many coefficients are estimated at the 0.05 significance level or better, with signs that align with prior expectations. People prefer safer and more luxurious cars, cars with more gears, heavier cars, and cars with higher ground clearance. Households prefer sedans over hatchbacks and SUVs; however, the preference for sedans decreases with family size, with larger families 14 The safety features of the vehicle include airbags, rear seatbelts, antilock brakes, and traction control. The luxury features include air-conditioning, power steering, central locking, power windows, alloy wheels, leather seats, power mirrors, and a CD player. 12

16 preferring SUVs over hatchbacks. People prefer more powerful cars (bigger engines) but smaller exterior dimensions, holding body type constant. The coefficients in Table 3 imply that, other things equal, driving distance increases with family size and decreases with age. Women drive less than men, and owners of automatics drive less than owners of manual transmission cars. Finally, driving distance is greater for safer, more luxurious cars. Table 4 displays the elasticities of fuel consumption implied by our model. We calculate elasticities by varying, for example, diesel fuel price holding petrol fuel price constant and allowing buyers to switch vehicles as well as distance driven when fuel price changes. Raising the price of diesel fuel by 5 percent lowers diesel fuel consumption by 7.2 percent, implying a long-run, own-price elasticity of consumption of Most of this reflects a shift from diesel to petrol cars; conditional on buying a diesel car, fuel consumption falls by 2.3 percent. 15 The corresponding long-run own-price elasticity of petrol is Our long-run, own-price elasticities are in line with studies in Europe (Graham and Glaister 2002; Verboven 2002). Graham and Glaister note that elasticities are higher in countries with higher fuel prices and in countries where consumers can substitute between petrol and diesel fuels. 17 The costs of diesel and petrol in India, while high at market exchange rates, are higher than in the UK and Europe when evaluated in purchasing power parity (PPP) terms. 18 The income elasticities in Table 4 are in line with the literature. Conditional on purchasing a diesel (petrol) vehicle, the income elasticity of fuel consumed is about Given that this is conditional on vehicle ownership, the appropriate comparison with the literature is with short-run elasticities. Graham and Glaister (2002) report short-run income elasticities of 15 The conditional elasticity of diesel fuel consumption is the elasticity of average fuel consumption by diesel car drivers when diesel price rises. 16 We note that these elasticities reflect the switch to vehicles of a different fuel type rather than a switch to more fuel-efficient vehicles. In baseline runs, the average fuel economy of petrol and diesel vehicles purchased differs little. Average fuel economy is insensitive to changes in diesel and petrol fuel prices. 17 We were unable to find published estimates of petrol and diesel fuel elasticities for India based on household data. Published estimates of petrol prices elasticities (e.g., Ramanathan 1999) are based on aggregate time-series data, which typically yield smaller elasticities (in absolute value) than estimates based on cross-sectional household data. 18 According to the Times of India (2011), India has the third-highest fuel prices, per liter, of any country in the world, when measured in PPP terms. In 2011, the prices per liter of diesel and petrol in India were $2.46 and $3.95 in PPP terms, compared with $1.91 and $1.85 in the UK. 19 The higher long-run income elasticity of diesel fuel consumption than that of petrol reflects the fact that an increase in income increases the market share of diesel vehicles. Recall that our model includes no outside good; hence our long-run elasticities are conditional on buying a car. 13

17 fuel consumption between 0.35 and Other studies have found low estimates of the impact of income on vehicle kilometers traveled (VKT), conditional on car ownership. 20 Using the 1990 Nationwide Personal Transportation Survey, Bento et al. (2005) find an elasticity of miles driven with respect to income of 0.12 for two-vehicle households and 0.23 for one-vehicle households. Studies by Mannering and Winston (1985) and Train (1986) also suggest that income has a small effect on distance driven, holding number of vehicles constant. Finally, we note the high elasticity of diesel fuel consumption with respect to the price of diesel cars ( 1.62). This suggests that a diesel car tax would be effective in moving buyers to petrol cars, although the impact on fuel consumption would be offset by increased petrol consumption. 6. Policy Simulations We use our behavioral model to explore the market and welfare implications of taxing diesel fuel and taxing diesel cars. We begin with a tax on diesel fuel that equates the prices of diesel and petrol, a policy considered by the Expert Group on a Viable and Sustainable System of Pricing of Petroleum Products (Parikh 2010). In 2010, diesel sold for 12.7 Rs/L less than petrol because of a lower rate of taxation. Policy 1 imposes a tax of 12.7 Rs/L on diesel, raising its price by 34.5 percent. 21 Policy 2 imposes a tax on new diesel vehicles. To make this policy comparable to policy 1, we set the diesel car tax at 24.7 percent, the rate that results in the same after-tax market share for diesel vehicles as the fuel tax. For the car tax to result in the same reduction in total fuel consumption would require a tax of over 80 percent on new diesel cars, which we view as politically infeasible. Because the cost of vehicle ownership is annualized in our econometric model, we annualize both the car tax charged to consumers and the revenues received by the government. Policy 3 examines an alternate diesel fuel tax of 2.77 Rs (7.55 percent), which results in the same total fuel conservation as the diesel car tax. 20 Changes in fuel prices have little impact on the fuel efficiency of vehicles purchased; thus our fuel consumption elasticities are approximately equal to VKT elasticities. 21 A natural question is how this tax compares with the value of externalities associated with fuel use. Parry et al. (2014) compute the value of externalities per liter of diesel and petrol associated with greenhouse gases, congestion, local air pollution, and accidents for countries throughout the world. If the value of a statistical life is transferred from the United States to India at market exchange rates, assuming an income elasticity of 1, the damages in 2010 are 12.4 Rs/L for petrol and 13.1 Rs/L for diesel used in passenger vehicles. See the appendix for details. 14

18 6.1. Results For each policy, we use the model of section 3 to compute the impact of the policy on market shares, driving distances, fuel consumption, consumer welfare, and government revenues. In computing fuel consumption, government revenues, and aggregate welfare impacts, we extrapolate results from our sample to all 2010 new car buyers. 22 Comparisons of market outcomes and welfare results under the policy simulations are presented in Table 5. For ease of comparison, we present model simulation results in the absence of any policies (prepolicy baseline). Policy 1. Equalizing Diesel and Petrol Fuel Prices As expected, the price-equalizing diesel fuel tax (Policy 1) has the greatest impact of the three policies on fuel consumption, reducing total fuel consumption by new car buyers by 6.99 percent. Under Policy 1, 66 percent of the reduction in diesel fuel use occurs because people switch to petrol vehicles: about 26 percent of diesel buyers in the prepolicy world switch to petrol vehicles after the tax (i.e., diesel market share falls from 33.6 to 24.8 percent). But this reduction in diesel consumption is largely offset by an increase in petrol consumption. The buyers who continue to purchase diesel cars decrease their fuel consumption by about 18 percent. They account for 75 percent of the reduction in total fuel consumption, which amounts to 7 percent of the fuel used by new car buyers. In reality, in 2010, the diesel fuel tax of 34.5 percent would have achieved much greater reductions in diesel fuel use in the used car market than in the new car market. Our simulations suggest that new car buyers who do not switch to petrol cars when the tax is imposed reduce their diesel fuel consumption by about 18 percent. A similar reduction in fuel consumption among existing diesel cars would be substantial, given that the stock of registered cars as of 2009 was 6.6 times the number of new cars sold in 2010 (Ministry of Road Transport and Highways 2012). 23 We estimate the welfare effects of the diesel tax by calculating compensating variation for new car owners in our dataset and extrapolating the results to the population of new car 22 In 2010 in India, 2,309,000 new passenger vehicles were purchased (SIAM, various years). 23 Our current dataset does not permit analysis of the impact of any of our policies on the stock of used cars, as we have data only on new car buyers. 15

19 owners. 24 The compensating variation associated with the diesel tax is, on average, 6,260 Rs per new car buyer, or about 14.4 billion Rs in the aggregate. The burden of the tax, however, falls entirely on would-be diesel car buyers. Persons who would not buy a diesel prior to the tax bear none of the tax, given that the price of petrol is not changed by the policy. Total compensating variation divided by the number of persons who bought a diesel in 2010 is 18,400 Rs, or about 3.0 percent of the average income of diesel buyers. The welfare effects of the tax increase as a percentage of income as income falls, given the low income elasticity of fuel consumption. The ultimate regressivity of the fuel tax depends on what is done with the increased fuel tax revenues. The diesel fuel tax generates more than 9.5 billion rupees in tax revenues. Because the initial diesel/petrol price difference (12.7 Rs) is due to the higher tax on petrol, the increase in petrol consumption generates additional tax revenues in the petrol market. 25 Subtracting both of these from compensating variation implies that the deadweight loss of the diesel tax is about 12 Rs per liter of fuel conserved. Policy 2. Taxing Diesel Cars By construction, the diesel car tax results in the same shift in car ownership from diesel to petrol vehicles as the fuel tax about 26 percent of diesel car buyers shift to petrol vehicles but their reduction in diesel fuel consumption is largely offset by increased consumption of petrol. On net, the 25 percent car tax results in a 2.0 percent reduction in total fuel consumption by new car buyers. The welfare cost of the car tax is lower, on average, than the fuel tax: compensating variation is about 3,370 Rs per household per year. The compensating variation per liter of fuel conserved (120 Rs/L) is, however, almost twice that of Policy 1 (64 Rs/L) because of the smaller impact of Policy 2 on fuel consumption. 24 For each random draw from the distribution of taste parameters, compensating variation is calculated by solving the equation ( ) ( ). The left-hand side of this equation is the indirect utility function in equation (1) without the term. The right-hand side is the counterfactual version of the left-hand side, with observed values of rental price, operating cost, and so on replaced with their postpolicy values, and with the income term augmented by compensating variation. Since indirect utility is not linear in income, the equation must be solved numerically. Expected compensating variation is then calculated by averaging over the results from each draw, and results are aggregated across individuals within a category (e.g., petrol car buyers). See, for example, Herriges and Kling (1999) for further discussion. 25 There is also a change in tax revenues from preexisting car taxes as buyers switch from diesel to petrol vehicles. All states levy taxes on the purchase price of a new car. Because diesel cars are, in general, more expensive than petrol cars, state tax revenues fall by approximately 1.1 billion Rs. We subtract this from the increase in fuel tax revenues in calculating deadweight loss. 16

20 The deadweight loss from the diesel car tax is actually negative. Revenues from increased sales of petrol, when added to revenues from the car tax, are greater than the amount that new car buyers must be compensated to restore them to their pretax level of utility. 26 The deadweight loss per liter of fuel saved is 27 Rs/L. Goulder and Williams (2003) note that ignoring the impact of a tax in one market on consumption in other markets with preexisting taxes can lead to biased estimates of the deadweight loss of a tax. This is clearly the case here. In the case of Policies 1 and 2, the shift of new car buyers to petrol vehicles increases petrol consumption by more than 13 percent (approximately 270 million liters), resulting in additional tax revenues of over 3 billion Rs. Policy 3. A Diesel Fuel Tax That Achieves the Same Fuel Conservation as the Car Tax Whether the diesel car tax is superior to the diesel fuel tax depends on the policymaker s objectives. If they are to reduce fuel consumption, the car tax is not very effective. A fairer comparison from the perspective of fuel conservation is between the diesel car tax and a tax on diesel fuel that would achieve the same reduction in total fuel consumption. A 7.6 percent (2.77 Rs) tax on diesel would also reduce total fuel consumption by 2.0 percent among new car buyers. The tax would reduce the market share of diesel vehicles from 34 percent to 31 percent, an 8.8 percent reduction. Over 60 percent of the reduction in diesel consumption that results from this tax (133 million liters per year) is due to the shift to petrol vehicles; however, this reduction is offset by an increase of 68 million liters in petrol consumption. The remainder of the reduction in diesel consumption is due to a 4.4 percent reduction in diesel consumption by buyers who continue to purchase diesel cars. 27 While compensating variation per liter of fuel conserved is only slightly lower than the larger diesel fuel tax (57.7 Rs/L), deadweight loss per liter of fuel conserved is smaller (2.13 Rs/L) Discussion The results above are subject to two caveats. The first is that, because we have data only on new car buyers, the model we estimate includes no outside good. Therefore, we cannot estimate the impact of the policies on the total number of cars sold. To the extent that both 26 In evaluating the deadweight loss of the car tax, we subtract the fall in state tax revenues (see footnote 25) from the increase in fuel tax revenues. 27 The reduction in diesel consumption by diesel car buyers accounts for approximately 76% of the total reduction in fuel consumed. 17

21 policies raise the cost of car ownership, they are likely to reduce new car sales. Our analysis above, because it ignores this impact, is likely to understate the reduction in fuel savings from both policies. The second caveat is that we do not estimate supply-side responses to our policies. 28 It is likely that automakers might react to either policy by lowering the price of diesel cars. This would attenuate the impact of all policies in shifting buyers to petrol vehicles and would reduce fuel savings. Both the diesel car tax and the diesel fuel tax would have impacts in the used car market. The diesel fuel tax would likely hasten the retirement of diesel (versus petrol) cars and would reduce fuel consumption by existing diesel vehicles. The tax on new diesel cars would likely increase the lifetimes of used diesel vehicles but would have no impact on the miles they are driven. We cannot analyze the welfare impacts of either policy on the used car market because of a lack of data; 29 however, ignoring the used car market clearly understates the fuel conservation benefits of the diesel fuel tax relative to a tax on new diesel cars. In a rapidly growing car market, the impacts of fuel conservation policies on future fuel consumption are arguably more important than impacts on the used car market. Taxing diesel fuel by 34 percent in 2010 would reduce fuel consumption by less than 300 million liters in 2010; however, keeping this policy in place would reduce fuel consumption by at least 40 times this amount in 2030, as a result of the rapid growth of the Indian car market. We base this statement on simple back-of-the envelope calculations. Using conservative assumptions about income growth, Arora et al. (2011) project new car sales in India from 2010 to 2030, estimating that sales will reach at least 9 million cars by Together with a survival curve for passenger vehicles, these projections yield estimates of the stock of vehicles purchased beginning in 2010 that will be on the road in years If vehicle miles traveled by new cars remain the same as in 2010, and if vehicle miles traveled decline by 2.5 percent each year as a car ages (Chugh et al. 2011), total fuel consumption by passenger vehicles in 2030 will be 40 times consumption levels in A diesel tax that reduces fuel consumption by 300 million liters today will, if kept 28 Estimating supply-side responses is difficult in the absence of an outside good. Specifically, own-price elasticities are likely to be underestimated in the absence of an outside good. 29 There is no survey of vehicle owners in India that is similar to the National Household Travel Survey in the United States. Data on used vehicle prices and odometer readings at the time of sale can be obtained online but do not include information on buyer characteristics. 18

22 in place, save 12 billion liters in 2030 alone. It will also save more than 4.8 billion liters over the life of 2010-vintage cars. These calculations underscore the benefits of enacting fuel conservation policies today Implications of Our Results for CAFE Standards Our model results also have important implications for fuel economy standards in India. The Bureau of Economic Efficiency is contemplating weight-based corporate average fuel economy standards that would go into effect in , with more stringent standards proposed for (Roychowdhury 2011). These would be weight-based standards, with heavier cars being allowed more liters per 100 km driven. Calculations based on our 2010 J. D. Power data suggest that the proposed standards would lower average liters per 100 km by 26 percent for petrol and 10 percent for diesel cars. 30 If the households in our sample were to buy the same vehicle after the increase in fuel economy and were to drive it the same number of kilometers, fuel consumption would decline by 26 percent for petrol and by 10 percent for diesel car owners. Weighting these percentages by market shares, total fuel consumption would decline by about 20 percent. The prediction of a 20 percent reduction in fuel consumption, however, ignores the impact of CAFE standards on vehicle choice and on kilometers driven. We use our model to estimate the impact of CAFE standards by increasing the liters per 100 km for each vehicle in the choice set according to the CAFE formula. This, correspondingly, reduces operating cost per kilometer driven. The results of our simulation are shown in Table 5. The greater improvement in the fuel economy of petrol (versus diesel) vehicles leads to a shift from diesel to petrol vehicles; indeed, we predict petrol market share to increase from 66 percent to 73 percent percent. One might expect this to increase fuel savings; however, the rebound effect in both petrol and diesel markets is sizable. The 26 percent average reduction in fuel cost per mile for petrol vehicles leads to a 22 percent increase in liters consumed (see Table 5). The 10 percent reduction in average diesel fuel costs per km results in a 5 percent increase in km driven for diesel vehicles. The net effect of these adjustments implies that the proposed CAFE standard would, once consumers adjust, reduce fuel consumption by 8.5 percent rather than 20 percent. 30 The standard reported by Roychowdhury (2011) is liters per km = *kerb weight (in kg) These figures correspond to a 36% increase in fuel economy for petrol cars and a 15% increase for diesel cars. 19

23 7. Conclusions The Indian government is taking steps to reduce the increased fuel consumption that is accompanying rapid motorization. Raising the tax on diesel fuel to equal the tax on petrol would be a natural way to achieve this; however, such a tax would have consequences that extend far beyond the passenger vehicle market. Only 5 percent of diesel fuel in India in consumed by passenger vehicles. Because of its use in agriculture and freight transport, diesel is widely regarded as the fuel of the poor. For this reason, increasing the tax on diesel fuel is politically difficult. We therefore compare a tax on new diesel cars, suggested by the Expert Group on a Viable and Sustainable System of Pricing of Petroleum Products (Parikh 2010) with a diesel fuel tax. The Expert Group called for an additional excise duty on a diesel vehicle corresponding to the differential tax on petrol. We interpret this as a car tax that has the same effect on diesel market share as a fuel tax that equates the prices of diesel and petrol. To analyze the effects of these policies, we estimate a structural econometric model of new car purchasing decisions and driving behavior. Our simulations suggest that in 2010, eliminating the petrol diesel price gap with a diesel fuel tax of 34 percent would reduce diesel market share by 26 percent. A diesel car tax of 25 percent would have the same impact on the market share of diesel vehicles. The car tax would reduce total (i.e., diesel plus petrol) fuel consumption in the new car market by 2.0 percent. The compensating variation associated with the tax would be approximately 120 Rs per liter of fuel conserved; however, the deadweight loss of the car tax is negative. This occurs because of the preexisting tax on petrol. As consumers who would have purchased diesel vehicles shift to petrol vehicles, petrol consumption increases by more than 12 percent (approximately 240 million liters), resulting in additional tax revenues of over 3.0 billion rupees. This must be subtracted from compensating variation to calculate the deadweight loss of the diesel car tax. The diesel fuel tax of 34 percent would achieve a greater reduction in total fuel consumption about 7 percent in the new car market by providing new car buyers who continue to buy diesel cars an incentive to drive fewer miles. These buyers reduce fuel consumption by about 18 percent and account for 75 percent of the total reduction in fuel consumption. The deadweight loss of the large diesel fuel tax is larger per liter of fuel saved than the car tax approximately 12 Rs per liter but the compensating variation per liter of fuel saved is much lower, approximately 64 Rs per liter. A fairer comparison might be between the car tax and a smaller tax on diesel fuel that would achieve the same reduction in fuel consumed in the new car market, a tax of 20

24 approximately 7.6 percent. This diesel tax also has a negative cost per liter of fuel saved and a welfare cost to consumers of approximately 58 Rs liter of fuel conserved. Both diesel fuel taxes would, however, result in additional fuel savings by used diesel vehicles, which would swamp any savings in the new car market, given the size of the diesel vehicle fleet. The diesel car tax, in contrast, would likely raise the price of used diesel cars but would provide no incentive to reduce kilometers traveled by diesel car owners. Our results also have important implications for fuel economy standards in India. Our estimates of the price elasticity of demand for fuel, conditional on owning a car of a given fuel type, are 0.8 for petrol and 0.6 for diesel fuel. 31 The high price elasticity of miles driven with respect to fuel price suggests that the fuel economy standards currently being contemplated by the government are likely to result in a substantial rebound effect. The Bureau of Economic Efficiency is contemplating weight-based corporate average fuel economy standards that would go into effect in , with more stringent standards proposed for (Roychowdhury 2011). These would be weight-based standards, with heavier cars being allowed more liters per 100 km driven. Calculations based on our 2010 J. D. Power data suggest that the proposed standards would lower average liters per 100 km by 26 percent for petrol and 10 percent for diesel cars. This reflects the fact that diesel cars are, on average, heavier than petrol cars. The greater improvement in fuel economy of petrol vehicles would likely lead to a shift from diesel to petrol vehicles; however, the rebound effect in both markets is substantial. Our estimates suggest that it is approximately 58 percent. This suggests that attention be paid to the possibility of raising diesel taxes. Diesel taxes would have repercussions throughout the economy; however, several countries tax diesel fuel for transport at a different rate than diesel fuel in other sectors (e.g., agriculture), and the same could be done in India. A recent study by Datta (2010) suggests that an economy-wide tax on diesel would not be regressive. Were this to be adopted, it would be a far more efficient way of reducing fuel consumption than other policies that the government is considering. 31 Long-run elasticities, which allow buyers to switch fuel type, are 1.27 for petrol and 1.44 for diesel. 21

25 Tables and Figures VARIABLES Table 1. Sales-Weighted Vehicle Summary Statistics UNITS Petrol hatchback Diesel hatchback Petrol sedan Diesel sedan Petrol SUV Price 10^5 rupees (2010) Diesel SUV (1.01) (0.755) (3.27) (2.59) (7.74) (3.76) Fuel economy kilometers/liter (1.06) (0.576) (0.828) (0.983) (0.280) (0.753) Operating cost 2010 rupees/kilometer (0.265) (0.0850) (0.276) (0.185) (0.0998) (0.193) Engine size cubic centimeters (0.160) (0.0664) (0.257) (0.210) (0.155) (0.186) Power ratio horsepower/kilogram ( ) ( ) (0.0124) (0.0107) ( ) (0.0185) Torque kilogram-meters (0.0180) (0.0295) (0.0326) (0.0457) (0.0165) (0.0514) Gears (0.268) (0) (0.117) (0.138) (0.149) (0) Automatic (0) (0) (0.127) (0.104) (0) (0.137) Length meters (0.173) (0.0844) (0.168) (0.144) (0.200) (0.152) Width meters (0.0775) ( ) (0.0376) (0.0418) (0.0515) (0.0588) Height meters (0.0748) (0.0479) (0.0515) (0.0331) (0.0179) (0.0862) Ground clearance meters (0.0531) (0.0556) (0.0876) (0.0706) (0.0373) (0.124) Weight 1,000 kilogram (0.118) (0.0415) (0.115) (0.112) (0.0941) (0.163) Safety index (0.597) (0.401) (0.779) (0.719) (0.298) (0.641) Luxury index (1.32) (0.865) (1.60) (1.25) (1.19) (1.90) # models Notes: This table presents weighted means with standard deviations in parentheses. Version level vehicle characteristics data come from AutoCar India and DriveInside.com. Model/fuel type level vehicle characteristics are constructed as the unweighted average across all available versions of each model/fuel type. The sales-weighted average of these is calculated for each vehicle category. Price and fuel economy data are averaged over all J. D. Power APEAL survey respondents that purchased each vehicle type. Luxury index is defined as the sum of the dummy variables for airconditioning, power steering, central locking, power windows, alloy wheels, leather seats, power mirrors, and CD player. Safety index is defined as the sum of the dummy variables for airbags, rear seatbelts, antilock braking system, and traction control. 22

26 Table 2. Demographic Summary Statistics VARIABLES UNITS Petrol hatchback Diesel hatchback Petrol sedan Diesel sedan Petrol SUV Income 10^5 rupees (2010) Diesel SUV (2.63) (2.56) (3.03) (2.86) (3.14) (2.95) Family size (1.55) (1.51) (1.63) (1.63) (1.56) (1.61) Age years (11.4) (10.7) (10.4) (9.18) (10.2) (9.84) % female (0.281) (0.179) (0.211) (0.1793) (0.123) (0.165) Driving distance kilometers/month 14,500 24,000 16,600 22,500 16,600 26,300 (16,000) (25,200) (18,00) (22,200) (16,800 (25,100) # observations 2, , Notes: This table presents unweighted means with standard deviations in parentheses. Owner demographics come from the 2010 J. D. Power APEAL survey. 23

27 Table 3. Demand Model Parameter Estimates FIXED Coefficient (Standard error) Hatchback * ( ) Sedan *** ( ) Age *** ( ) Female 0.293*** (0.0585) Family size *** (0.0101) Engine size *** ( ) Power ratio (0.0627) Torque (0.0112) Length *** ( ) Width *** ( ) Height *** ( ) Ground clearance *** ( ) Weight ** ( ) Gears *** ( ) Automatic *** ( ) Safety index *** ( ) Luxury index *** ( ) Family size hatchback *** ( ) Family size sedan *** ( ) Age safety index 2.29E-05 (3.03E-05) Family size power ratio ( ) Scale factor (µ) 0.910*** (0.0776) Taste for driving (σ) 0.859*** (0.0225) RANDOM Coefficient (Standard error) Standard deviation (Standard error) Income rent (β) 2.87*** (0.0977) (0.0322) Operating cost (α) 1.44*** (0.107) 0.441*** (0.0551) Notes: This table presents full information maximum likelihood coefficient estimates with 12 make fixed effects (not shown). Integrals simulated using 200 shifted and shuffled Halton draws. Number of observations = 6132, LL = 30,496 at convergence. *p < 10%. **p < 5%, ***p < 1%, based on simulation noise robust standard errors. 24

28 Table 4. Fuel Consumption Elasticities Petrol (long run) Petrol (conditional) Diesel (long run) Diesel (conditional) Own-price (fuel) Cross-price (fuel) Income Own-price (car) Cross-price (car) Notes: All elasticities are calculated by varying price or income by 5% from baseline values, allowing households to vary vehicle choices and miles driven. Long-run elasticities are based on changes in total fuel consumption by all households and reflect changes in the type of vehicle bought. Conditional elasticities are based on change in average fuel consumption, conditional on buying a vehicle of a particular fuel type. 25

29 Table 5. Policy Simulation Results (POLICY 1) (POLICY 2) (POLICY 3) Market share Prepolicy baseline 34.46% diesel fuel tax 24.65% diesel car tax 7.54% diesel fuel tax Fuel economy standards petrol 66.4% 75.2% 75.2% 68.6% 72.9% diesel 33.6% 24.8% 24.8% 31.4% 27.1% Average fuel consumption petrol 1,280 1,290 1,290 1,290 1,190 (L/year) diesel 1,610 1,310 1,600 1,540 1,560 Average fuel economy petrol (km/l) diesel Total fuel consumption petrol 1,970 2,240 2,230 2,040 1,190 (10 6 L/year) diesel 1, ,110 1,560 Total fuel conserved (10 6 L/year) Total compensating variation (CV) (10 6 Rs 2010/year) 14,400 7,780 3,790 CV/liter conserved (Rs 2010/L) Δ government revenue (10 6 Rs 2010/year) 11,800 9,560 3,650 Deadweight loss/liter (Rs 2010/L) Notes: This table presents policy simulation results for year 2010 using parameter estimates presented in Table 3. The 2010 petrol and diesel fuel prices (in 2010 Rs/liter) were and 36.72, respectively. Thus a 7.55% diesel fuel tax amounts to 2.77 Rs, and a 34.46% diesel fuel tax amounts to Rs. In both cases, the existing tax on petrol fuel is assumed to equal Rs. 26

30 Figure 1. Nominal Fuel Price (Rs/Liter) for Petrol and Diesel, Source: Ministry of Petroleum and Natural Gas, Government of India 27

31 Figure 2. Dieselization across Passenger Vehicle Segments, Source: SIAM (various years) 28

32 Figure 3. Model Fit (Market Shares) Source: Authors calculations. 29

33 Figure 4. Model Fit (Driving Distance) Source: Authors calculations. 30